1 ; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI-SAFE -check-prefix=SI -check-prefix=FUNC %s
2 ; RUN: llc -march=amdgcn -enable-no-nans-fp-math -verify-machineinstrs < %s | FileCheck -check-prefix=SI-NONAN -check-prefix=SI -check-prefix=FUNC %s
3 ; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
5 declare float @llvm.fabs.f32(float) #1
6 declare float @llvm.floor.f32(float) #1
8 ; FUNC-LABEL: {{^}}cvt_flr_i32_f32_0:
9 ; SI-SAFE-NOT: v_cvt_flr_i32_f32
11 ; SI-NONAN: v_cvt_flr_i32_f32_e32 v{{[0-9]+}}, s{{[0-9]+}}
13 define amdgpu_kernel void @cvt_flr_i32_f32_0(i32 addrspace(1)* %out, float %x) #0 {
14 %floor = call float @llvm.floor.f32(float %x) #1
15 %cvt = fptosi float %floor to i32
16 store i32 %cvt, i32 addrspace(1)* %out
20 ; FUNC-LABEL: {{^}}cvt_flr_i32_f32_1:
21 ; SI: v_add_f32_e64 [[TMP:v[0-9]+]], s{{[0-9]+}}, 1.0
22 ; SI-SAFE-NOT: v_cvt_flr_i32_f32
23 ; SI-NONAN: v_cvt_flr_i32_f32_e32 v{{[0-9]+}}, [[TMP]]
25 define amdgpu_kernel void @cvt_flr_i32_f32_1(i32 addrspace(1)* %out, float %x) #0 {
26 %fadd = fadd float %x, 1.0
27 %floor = call float @llvm.floor.f32(float %fadd) #1
28 %cvt = fptosi float %floor to i32
29 store i32 %cvt, i32 addrspace(1)* %out
33 ; FUNC-LABEL: {{^}}cvt_flr_i32_f32_fabs:
35 ; SI-SAFE-NOT: v_cvt_flr_i32_f32
36 ; SI-NONAN: v_cvt_flr_i32_f32_e64 v{{[0-9]+}}, |s{{[0-9]+}}|
38 define amdgpu_kernel void @cvt_flr_i32_f32_fabs(i32 addrspace(1)* %out, float %x) #0 {
39 %x.fabs = call float @llvm.fabs.f32(float %x) #1
40 %floor = call float @llvm.floor.f32(float %x.fabs) #1
41 %cvt = fptosi float %floor to i32
42 store i32 %cvt, i32 addrspace(1)* %out
46 ; FUNC-LABEL: {{^}}cvt_flr_i32_f32_fneg:
48 ; SI-SAFE-NOT: v_cvt_flr_i32_f32
49 ; SI-NONAN: v_cvt_flr_i32_f32_e64 v{{[0-9]+}}, -s{{[0-9]+}}
51 define amdgpu_kernel void @cvt_flr_i32_f32_fneg(i32 addrspace(1)* %out, float %x) #0 {
52 %x.fneg = fsub float -0.000000e+00, %x
53 %floor = call float @llvm.floor.f32(float %x.fneg) #1
54 %cvt = fptosi float %floor to i32
55 store i32 %cvt, i32 addrspace(1)* %out
59 ; FUNC-LABEL: {{^}}cvt_flr_i32_f32_fabs_fneg:
61 ; SI-SAFE-NOT: v_cvt_flr_i32_f32
62 ; SI-NONAN: v_cvt_flr_i32_f32_e64 v{{[0-9]+}}, -|s{{[0-9]+}}|
64 define amdgpu_kernel void @cvt_flr_i32_f32_fabs_fneg(i32 addrspace(1)* %out, float %x) #0 {
65 %x.fabs = call float @llvm.fabs.f32(float %x) #1
66 %x.fabs.fneg = fsub float -0.000000e+00, %x.fabs
67 %floor = call float @llvm.floor.f32(float %x.fabs.fneg) #1
68 %cvt = fptosi float %floor to i32
69 store i32 %cvt, i32 addrspace(1)* %out
73 ; FUNC-LABEL: {{^}}no_cvt_flr_i32_f32_0:
74 ; SI-NOT: v_cvt_flr_i32_f32
76 ; SI: v_cvt_u32_f32_e32
78 define amdgpu_kernel void @no_cvt_flr_i32_f32_0(i32 addrspace(1)* %out, float %x) #0 {
79 %floor = call float @llvm.floor.f32(float %x) #1
80 %cvt = fptoui float %floor to i32
81 store i32 %cvt, i32 addrspace(1)* %out
85 attributes #0 = { nounwind }
86 attributes #1 = { nounwind readnone }